-
[show abstract]
[hide abstract]
ABSTRACT: A custom CMOS imager with integrated motion computation is described. The architecture is based on correlating in time moving edges. Edges are located in time by a custom sensor; and correlated in a coprocessing module. The sensor architecture is centered around a compact pixel with analog signal processing and digital self-signaling capabilities. The sensor pixels detect moving edges in the image and communicate their position using an address-event protocol associated to temporal stamps. The coprocessing module correlates the edges and computes the velocity vector map. The motion sensor could be used in applications such as self-guided vehicles, mobile robotics and smart surveillance systems. The article details the motion sensor architecture, the simulated performance, the VLSI implementation and some preliminary results on fabricated prototypes
Computer Architecture for Machine Perception, 1997. CAMP '97. Proceedings Fourth IEEE International Workshop on; 11/1997
-
[show abstract]
[hide abstract]
ABSTRACT: This paper presents a special-purpose VLSI architecture for dominant point extraction along 2D contours. Such dominant points carry useful information for shape analysis and pattern recognition applications since they represent a local shape property and segment object contours into piecewise linear segments and circular arcs. The proposed architecture implements an algorithm based on the curvature primal sketch. It consists of a set of 1D systolic FIR filters performing a multiresolution analysis of the scene's object contours, a set of finite-state-machines extracting zero-crossings and extrema of the filtered data, and a set of scale-space integration cells combining the accurate locations provided by the finest filters with the noise rejection properties of the coarsest ones in order to reliably extract relevant dominant points with accurate localization. The overall architecture has been successfully implemented and integrated to a custom machine vision system with real-time edge-extraction and edge-tracking capabilities. Some experimental results obtained using this system are presented and discussed. Performance issues are also addressed
Computer Architecture for Machine Perception, 1997. CAMP '97. Proceedings Fourth IEEE International Workshop on; 11/1997
-
[show abstract]
[hide abstract]
ABSTRACT: A goal of computer vision is the construction of scene
descriptions based on information extracted from one or more 20 images.
A reconstruction strategy based an a three-level representational
framework is proposed. The first representational level, the primal
sketch, makes explicit physical characteristics of the scene through
detection of illuminance changes and their geometrical distribution and
organization. Physical characteristics appear at several spatial scales
and a multiresolution analysis helps in eliminating spurious edges. The
second representational level, the raw 2.50 sketch, makes explicit the
orientation and rough depth at edge location of the visible surfaces. A
multiresolution neural network stereo algorithm is designed to compute
the disparity at each edge location and at all the resolution levels.
Matching is facilitated by a hierarchical focusing mechanism. The third
representation level, the full 2.50 sketch, makes explicit the
orientation and depth estimate at all the visible surface coordinates.
Depth information between the edges is computed with a local
shape-from-shading algorithm. A constraint satisfaction network fuses
stereo and shading data
Neural Networks,1997., International Conference on; 07/1997
-
[show abstract]
[hide abstract]
ABSTRACT: The exploitation of analog VLSI techniques combined with computer vision knowledge offers spectacular possibilities. Limitations of current VLSI technologies do not allow to create sensors with extremely complex pixel architecture, but the coupling of external CMOS analog processing units is a great solution for rapid low level segmentation processes. This paper presents a novel sensing approach where photo-transduction, multiresolution feature extraction, scale-space integration, and edge tracking combined with sub-pixel interpolation are performed on a mixed-signal (digital-analog) VLSI architecture. The paper also discusses how we implement the curvature primal sketch into the system for higher level scene representation. The main sensory part of this integrated image acquisition system is a CMOS sensor called Multiport Access photo-Receptor (MAR). VLSI also provides means to integrate analog computing, digital controller, and DSP co-processor modules which define a powerful sensory chip set for focal plane image processing. A current version of the MAR sensor which implements 256×256 pixels includes 16 analog spatial filters which simultaneously compute multiresolution edge maps. This novel smart image sensor approach with associated low level segmentation capability presents good opportunities for real time automated process for the particular case of unstructured environment
Computer Architectures for Machine Perception, 1995. Proceedings. CAMP '95; 10/1995
-
[show abstract]
[hide abstract]
ABSTRACT: This paper is concerned with the reliability of a shape
description recovered from a set of scattered measurements. The
recovering process of the description should not introduce a bias that
is caused by specific acquisition conditions such as the presence of
spurious measurements and the relative position of the sensor with
respect to the object. Moreover, the recovered description should be
stable for a sampling variation. While the fitting stage is based on a
measurement error model which takes into account the sensor's viewpoint,
the stability with sampling is tested by perturbing an hypothesized
section. The validity of the approach is demonstrated by extracting
reliable estimates of polynomial sections (lines, conics) from surface
profile range data obtained from one or several viewpoints
Pattern Recognition, 1994. Vol. 1 - Conference A: Computer Vision & Image Processing., Proceedings of the 12th IAPR International Conference on; 11/1994
-
[show abstract]
[hide abstract]
ABSTRACT: This paper presents a sensing approach where photo-transduction,
multi-resolution feature extraction, scale-space integration and edge
tracking are performed on a mixed (digital-analog) VLSI architecture in
order to generate medium-level scene description. The proposed system is
mainly targeted for robot vision applications where feature description
is preferred to a set of raw or raster 2D images and edge maps. The
Multiport Access photo-Receptor (MAR) is a CMOS sensor and represents
the main sensory part of this integrated image acquisition system. VLSI
also provides means to integrate analog computing, digital controller
and DSP co-processor modules which define a powerful sensory chip set
for focal plane image processing. A current version of the MAR sensor
which implements 256×256 pixels includes 16 analog spatial filters
which simultaneously compute multiresolution edge maps. This unique 2D
hexagonal smart sensor approach which performs up to 8.5×10<sup>9
</sup> arithmetic Op/sec during the acquisition/filtering phase and
25×10<sup>9</sup> Logical Op/sec for scale-space integration
allows high resolution image capability. It represents a significant
improvement for passive sensory units in a compact assembly for computer
vision applications
Computer Vision and Pattern Recognition, 1994. Proceedings CVPR '94., 1994 IEEE Computer Society Conference on; 07/1994
-
[show abstract]
[hide abstract]
ABSTRACT: Robust extraction of surface parameters from multiple view
scattered and noisy 3-D measurements is a delicate task. It is shown
that a stable local surface description can be extracted on sections
where measurement constraints are redundant with respect to a polynomial
model. A segmentation approach is developed to identify these sections.
The approach is based on a measurement error model which takes into
account the sensor's viewpoint. An application of the approach to the
extraction of straight line sections from single scan 3-D surface
profiles is presented
Computer Vision and Pattern Recognition, 1993. Proceedings CVPR '93., 1993 IEEE Computer Society Conference on; 07/1993
-
[show abstract]
[hide abstract]
ABSTRACT: The authors present a multimodule focal plane processing sensor
architecture which provides high-resolution (up to 512 × 512
pixels) multiscale real-time analog edge extraction for robot vision.
The hexagonal CMOS sensor uses a multiport addressing architecture of
the pixel array to apply external multiscale analog spatial convolution
followed by edge detection. The sensor architecture and its peripheral
analog filtering modules are described, and results obtained from a 256
× 256 prototype are presented. This analog satellite processing
approach may be extended to various types of computational sensors,
including a 3-D rangefinder, motion sensors, and tactile perception
devices
Custom Integrated Circuits Conference, 1993., Proceedings of the IEEE 1993; 06/1993
-
[show abstract]
[hide abstract]
ABSTRACT: A 3D camera based on the differential absorption of light by a
colored liquid is described. The range image is obtained from two
illuminance images taken at two different wavelengths. The theoretical
aspects of range-from-absorption are discussed in detail. Practical
considerations for the calibration and implementation of the method are
also covered. The accuracy of the 3D camera is discussed, and
experimental results are presented
IEEE Transactions on Instrumentation and Measurement 11/1992; · 1.21 Impact Factor
-
Intelligent Robots and Systems, 1992., Proceedings of the 1992 lEEE/RSJ International Conference on; 08/1992
-
[show abstract]
[hide abstract]
ABSTRACT: The neural network implementation of some commonly used edge
detectors is reviewed and compared. Edge detection is scale-dependent.
Edges are visible only over a range of scales. Multiple scale analysis
of the input image is required to have a complete description of the
edges. The authors propose a compact pyramidal multi-level neural net
architecture for image representation at multiple spatial scales.
Lateral weighted links within a level compute edge localization and
intensity gradient. Feedback between successive levels is used to
reinforce and refine the position of true edges
Neural Networks, 1992. IJCNN., International Joint Conference on; 07/1992
-
[show abstract]
[hide abstract]
ABSTRACT: The authors present a computer vision technique for the acquisition and processing of 3-D images of the profile of wax dental imprints in the automation of diagnosis in orthodontics. The acquisition of the 3-D images is based on the absorption of light by a dispersive medium and uses standard CCD (charge coupled device) cameras. The profiles of both sides of the imprint are acquired simultaneously. The 3-D image of each side of the imprint is segmented by nonlinear filtering of the 3-D data, and the interstices between the teeth are detected. Two operators are presented: one for the detection of the interstices between the teeth for incisors, canines, and premolars, and one for those between molars. A method for deciding the optimal neighborhood of application of each operator is also presented. Experimental results show that the two operators are very effective at detecting the interstices.
IEEE Transactions on Medical Imaging 02/1991; 10(3):453-61. · 3.64 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: The Multiport Array Receptor (MAR), a system which combines
optical sensing with integrated focal-plane processing capabilities, is
described. Its central element is a photosensor array with hexagonal
tesselation and complex peripheral selection logic which provides
parallel analog readout over prescribed areas. An external computing
module performs real-time spatial convolution at multiple resolutions
while a closed-loop microprogrammed controller addresses regions of
interest and supervises communication between the camera and the host
computer. This integrated image sensor and processor implements
programmed sequences of instruction primitives and yields a complete
state description of each processed pixel. It is capable of automatic
edge tracking and returns lists of connected pixels
Pattern Recognition, 1990. Proceedings., 10th International Conference on; 07/1990
-
[show abstract]
[hide abstract]
ABSTRACT: A practical algorithm for subsurface defect extraction (blobs) in
infrared or in low spatial content images is proposed. It first locates
potential defects by a spatial sorting of the pixels, placed in
decreasing order, based on their brightness. Labeling of the pixels is
based on the distance. Defect shape is grown by gradually decreasing the
threshold until a sudden increase in the number of pixels agglomerates
together or an image boundary is encountered
IEEE Transactions on Systems Man and Cybernetics 06/1990;
-
[show abstract]
[hide abstract]
ABSTRACT: An algorithm for extracting edges and plane regions of a polyhedral object in a three-dimensional (3D) range image is described. The object may be Convex or nonconvex. A model of the object is built with the regions extracted. Possible extension to cylindrical objects is also considered. The range images are obtained with a novel range-finder camera that can produce 128 × 256 or 256 × 256 surface element (surfcels) images. The edge detection is accomplished in five steps and yields edges one surfcel wide. The region-finding algorithm relies on the concept of the "hemispheric histogram." The histogram is built with the normals of groups of surfcels (patches) forming the image. Analysis of the hemispheric histogram gives global information on the surface orientation of the visible regions of an object. Once these regions are extracted, they are expanded with a region growing process. Geometric properties of the regions are computed by a simple contour following algorithm. Then, a relational model of the regions is built. The model gathers information that is independent of the position and orientation of the object ill the reference plane and could be Used for object recognition in an unsupervised 3D vision system.
IEEE Journal on Robotics and Automation 11/1987;
-
[show abstract]
[hide abstract]
ABSTRACT: It is possible to design amplifiers with virtually no input capacitance. As shown in the sequel, this feature, however, is not by itself sufficient to insure faithful recording of fast transmembrane potentials with glass microelectrodes.
IEEE Transactions on Biomedical Engineering 02/1976; · 2.28 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Presents a method to compute the inter-frame transformation
between two range image views of complex multi-part objects. No exact
feature matching is attempted and no initial approximate transformation
is provided. The method is naturally decomposed into two stages of
initial estimation and final refinement of the transformation. A
hierarchical triangulation-based surface representation provides an
efficient way to select the control points at which the alignment of the
two surfaces is to be evaluated. This representation also permits the
selection of a manageable number of initial transformations among which
at least one is to be in the parametric neighborhood of the actual
transformation. Experimental results show that the computed
transformation between two views of a complex multi-part object may
provide angles of rotation within a fraction of a degree of the actual
ones
Pattern Recognition, 1992. Vol.I. Conference A: Computer Vision and Applications, Proceedings., 11th IAPR International Conference on;
-
[show abstract]
[hide abstract]
ABSTRACT: We present a method to compute the interframe transformation between two range image views of complex multipart objects. No exact feature matching is attempted and no initial approximate transformation is provided. The method is naturally decomposed into two stages of initial estimation and final refinement of the transformation. A hierarchical triangulation-based surface representation provides an efficient way to select the control points at which the alignment of the two surfaces is to be evaluated. This representation also permits the selection of a manageable number of initial transformations among which at least one is to be in the parametric neighborhood of the actual transformation. Previous techniques are compared and their adaptation into an integrated method makes it possible to go beyond the identified limitations. Experimental results show that the computed transformation between two views of a complex multipart object may provide angles of rotation within a fraction of a degree of the actual ones.
Computer Vision and Image Understanding.
